Audacity normalize volume: Normalize, Amplify or Loudness Normalization

Under Effect, Volume and Compression you will find Amplify, Normalize and Loudness Normalization. All three change how loud the audio is. They answer different questions, and picking the wrong one is why a recording can measure correctly and still sound quiet.

Amplify moves everything up or down by an amount you choose. It is a volume knob applied to the file.

Normalize raises the audio until its single loudest peak hits a level you set, usually minus 1 dB. It is peak-based.

Loudness Normalization targets perceived loudness measured in LUFS, which is much closer to how loud something actually sounds to a person. For anything you are publishing, this is usually the one you want.

Imagine a forty-minute episode recorded at a comfortable level, with one moment where someone knocks the desk. That knock is the loudest peak in the file.

Normalize sets the level by that peak. It raises the file until the knock reaches minus 1 dB, which means the forty minutes of speech underneath barely move. The file is technically normalised and still sounds quiet, because one irrelevant transient decided the outcome.

This is the single most common reason people normalise a podcast and then wonder why it is quieter than every other show in the feed. Peak normalisation measures the wrong thing for speech.

LUFS measures loudness the way a listener experiences it, averaged over the whole programme rather than caught at a single instant. Streaming platforms use it to decide whether to turn your audio up or down, which they all now do automatically.

Common targets: around minus 16 LUFS for stereo podcast audio, around minus 19 LUFS for mono, and around minus 14 LUFS for YouTube and most music streaming. Apple Podcasts and Spotify both normalise towards roughly minus 16, so delivering close to that means nothing gets adjusted on the way in.

In Audacity, select the whole track, open Loudness Normalization, set perceived loudness to your target, and apply. Leave the option to normalise stereo channels independently switched off for speech, because it will shift a voice around the stereo image.

Delivering louder than the target gains you nothing. The platform simply turns it down, and you have given away dynamic range for no benefit.

Amplify is the tool when you want to change one part of a recording rather than all of it. Select a passage where someone turned away from the microphone, apply a few decibels, and move on.

It also shows you the headroom you have. Open Amplify on a selection and the New Peak Amplitude field tells you how close to clipping you already are, which is a quick way to check a recording before you do anything else.

Leave the allow clipping box unticked. There is no situation in spoken audio where deliberately clipping is the answer.

Loudness normalisation belongs at the very end: after noise reduction, after EQ, after compression, after editing. Every one of those steps changes the level, so normalising early means doing it again anyway.

Normalising twice is worth avoiding. Each pass is a calculation on already-processed audio, and repeated rounds accumulate small errors for no gain.

If different speakers in one episode are at very different levels, fix that with compression or by amplifying individual sections first. Loudness normalisation works on the whole file and cannot even out the relationship between two voices.

Hilite sets publish-ready loudness automatically at export, so the LUFS target is not something you have to think about per episode. We make it, so weigh that accordingly. Audacity's Loudness Normalization does the same job for free once you know to use it instead of Normalize.